id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
e5a7783e-1bcd-48d0-847f-01b552826b0b | StampyAI/alignment-research-dataset/blogs | Blogs | fuzzies & utils: check that you're getting either
fuzzies & utils: check that you're getting either
-------------------------------------------------
the post [*purchase fuzzies and utilons separately*](https://www.lesswrong.com/posts/3p3CYauiX8oLjmwRF/purchase-fuzzies-and-utilons-separately) puts forward the idea t... |
42addbb6-e625-45fd-854c-626f94590a09 | trentmkelly/LessWrong-43k | LessWrong | Questions on SI Research
Hello LessWrong,
As one of my assignments at the Singularity Institute (SI), I am writing a research FAQ answering the most frequently asked questions regarding the Singularity Institute's research program.
For a short summary of what SI is about, see our concise summary.
Here are some ex... |
f52f520f-ed76-4188-8819-c7fee997e7ce | trentmkelly/LessWrong-43k | LessWrong | Can Reasoning Models Avoid the Most Forbidden Technique?
In a recent post, Zvi described what he calls "The Most Forbidden Technique":
> An AI produces a final output [X] via some method [M]. You can analyze [M] using technique [T], to learn what the AI is up to. You could train on that. Never do that.
>
> You trai... |
c21dc10f-789c-4646-83ec-ac1ac297bf25 | trentmkelly/LessWrong-43k | LessWrong | Some Summaries of Agent Foundations Work
This is a bunch of not-very-technical summaries of agent foundations work from LessWrong and the Alignment Forum.
I was hoping to turn it into a comprehensive literature review, categorising things in an enlightening way and listing open problems. It turns out that's pretty ha... |
d4810c76-9376-4462-a9ff-30a30e43ed7d | trentmkelly/LessWrong-43k | LessWrong | As a Washed Up Former Data Scientist and Machine Learning Researcher What Direction Should I Go In Now?
Some background, I was interested in AI before the hype, back when neural networks were just an impractical curiosity in our textbooks. I went through an undergrad in Cognitive Science and decided that there was so... |
bb238eb9-8121-4f63-9efa-46b5473d7153 | trentmkelly/LessWrong-43k | LessWrong | Feeling Old: Leaving your 20s in the 2020s
A younger friend recently asked why I (and many others my age) make so many comments about being old. After some thought, I realized a big part of this was related to experiencing specific meaningful life stages during the pandemic.
In 2019, I could (and often did) stay up l... |
de6aae5c-6aee-438c-b0fa-490c9d45023e | trentmkelly/LessWrong-43k | LessWrong | Progress links digest, 2023-10-12: Dyson sphere thermodynamics and a cure for cavities
I’ve been traveling for a while, so this is a long one, covering the last ~month. I tried to cut it down, but there have been so many amazing announcements, opportunities, etc.! Feel free to skim and jump around:
From the Roots of... |
558f9a22-a180-44cb-a0c7-2119a69bb73b | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | What does the launch of x.ai mean for AI Safety?
[x.ai](https://x.ai/) has now launched. It seems worthwhile to discuss both what it means for AI safety and whether people interested in AI safety should [consider applying](https://airtable.com/appO5m0cwztccTdiA/shrtrQqaaoBt4Xm8i) for the company.
**Some thoughts:**
... |
76cbba55-ca25-40a9-a597-3610b0eac51c | trentmkelly/LessWrong-43k | LessWrong | Open Thread July 2018
For those low-effort ideas that are worth a little attention but perhaps not a lot (yet).
(shouldn't a bot be posting these?)
Prompts for things to say:
* What accomplishments are you celebrating from the last month?
* What are you reading?
* What subject has taken up most of your thoughts ... |
07330600-95c1-4d43-90b7-06c2bac8eec3 | StampyAI/alignment-research-dataset/special_docs | Other | Intelligence Explosion Microeconomics
MIRI
MACHINE INTELLIGENCE
RESEARCH INSTITUTEIntelligence Explosion Microeconomics
Eliezer Yudkowsky
Machine Intelligence Research Institute
Abstract
I. J. Good’s thesis of the “intelligence explosion” states that a sufficiently advanced ma-
chine intelligence could build a smarter v... |
cab1d86d-12bd-4b08-9984-cf2e76d08eb0 | trentmkelly/LessWrong-43k | LessWrong | An Update on Academia vs. Industry (one year into my faculty job)
I've been an assistant professor (equivalent) for ~1 year now at Cambridge. Shortly after accepting the position, I wrote AI x-risk reduction: why I chose academia over industry.
Since then, I've had a lot of conversations on academia vs. industry wi... |
8e5a9011-c1a8-4da3-9ebf-38313fc80b08 | trentmkelly/LessWrong-43k | LessWrong | Kodo and Din
I got these words from a Duncan Sabien lecture and kept wanting to link to them. Since Duncan hasn’t written them up as an essay yet, I’m doing it with permission; I’ll update with a link to his version if he ever writes it. [Update: He wrote it. It should be free to read by early May.]
I think it’s trad... |
329c62ec-eed4-4be3-89a7-de87ed0d360d | trentmkelly/LessWrong-43k | LessWrong | Meetup : Moscow meetup: The Toulmin model of argument, belief investigation exercise, psycholinguistics
Discussion article for the meetup : Moscow meetup: The Toulmin model of argument, belief investigation exercise, psycholinguistics
WHEN: 15 February 2015 02:00:00PM (+0300)
WHERE: Russia, Moscow, ulitsa L'va Tolst... |
1a242d9d-37ff-4ac4-bffe-1ce4b82fb3e4 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | How transparency changed over time
*This post is written as part of the SERI MATS program.*
This post explores how model interpretability has changed over time, and if the existing approaches to interpretability are likely to remain viable as the models increase in size and capability. I examine several approaches to... |
cebae89f-0e9a-4a3b-b673-33c51f1f79f4 | trentmkelly/LessWrong-43k | LessWrong | Towards Quantitative AI Risk Management
Reading guidelines: If you are short on time, just read the section “The importance of quantitative risk tolerance & how to turn it into actionable signals”
Tl;dr: We have recently published an AI risk management framework. This framework draws from both existing risk managemen... |
59e06e3f-6edf-4427-82e9-f9db838743aa | trentmkelly/LessWrong-43k | LessWrong | Self-Indication Assumption - Still Doomed
I recently posted a discussion article on the Doomsday Argument (DA) and Strong Self-Sampling Assumption. See http://lesswrong.com/lw/9im/doomsday_argument_with_strong_selfsampling/
This new post is related to another part of the literature concerning the Doomsday Argument - ... |
6af19c4d-d998-48b9-8143-cc607a255ea3 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Ranked Reward: Enabling Self-Play Reinforcement Learning for Combinatorial Optimization
1 Introduction and Motivation
------------------------------
Reinforcement learning (RL) algorithms that combine neural networks and tree search have delivered outstanding successes in two-player games such as Go, Chess, Shogi, ... |
468b24f0-88f6-4cb4-9843-fbac5a516809 | trentmkelly/LessWrong-43k | LessWrong | Does SI make everyone look like swimsuit models?
William Easterly believes Sports Illustrated’s swimsuit issue externalises toward women with their ‘relentless marketing of a “swimsuit” young female body type as sex object’. He doesn’t explain how this would happen.
As far as I can tell, the presumed effect is that p... |
68e94606-ab56-4a34-81b0-f76bab4e5217 | trentmkelly/LessWrong-43k | LessWrong | What is our current best infohazard policy for AGI (safety) research?
I struggle to see what the infohazard policy for AGI safety research writing should be. For example, it has been pointed out multiple times that GPT-3 paper was an infohazard, because so many people realised this kind of model was possible, and deve... |
72c8d0ff-d520-4f11-a4e2-25a4fb23fe67 | trentmkelly/LessWrong-43k | LessWrong | Making deals with AIs: A tournament experiment with
a bounty
Kathleen Finlinson & Ben West, May 2025
Mission
* We are working to establish the possibility of cooperation between humans and agentic AIs. We are starting with a simple version of making deals with current AIs, and plan to iterate based on what we lear... |
df95a503-a6a1-4cfd-b769-91ec5a70e6cd | trentmkelly/LessWrong-43k | LessWrong | New scorecard evaluating AI companies on safety
The new scorecard is on my website, AI Lab Watch. This replaces my old scorecard. I redid the content from scratch; it's now up-to-date and higher-quality. I'm also happy with the scorecard's structure: you can click on rows, columns, and cells and zoom in to various thi... |
6a9983c1-87c5-47bb-bd4b-76bfd6d7507e | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Retrospective on the 2022 Conjecture AI Discussions
At the end of 2022, following the success of the [2021 MIRI Conversations](https://www.lesswrong.com/s/n945eovrA3oDueqtq), [Conjecture](https://conjecture.dev/) started a project to host discussions about AGI and alignment with key people in the field. The goal was s... |
670d5e63-3c1b-4440-95f2-652c3da21e8f | trentmkelly/LessWrong-43k | LessWrong | Pecking Order and Flight Leadership
It was recently pointed out to me that humans are weird, compared to other social animals, in that we conflate the pecking order with the group decision-making process.
The pecking order, for instance in birds, is literally the ranking of who gets to eat first when food is scarce.
... |
b785281b-fcb1-42e2-871f-dbc1ad9fcdcf | trentmkelly/LessWrong-43k | LessWrong | Is OpenAI losing money on each request?
While working on another post, I decided to follow up some details by doing some naive modeling of OpenAI’s LLM API revenue stream. The naive approach seems inadequate, because it implies OpenAI requires many years to break even just on the cost of GPUs.
* OpenAI charges the f... |
78225d07-b964-46b9-9598-92e10b4ac31f | trentmkelly/LessWrong-43k | LessWrong | Making Fun of Things is Easy
Making fun of things is actually really easy if you try even a little bit. Nearly anything can be made fun of, and in practice nearly anything is made fun of. This is concerning for several reasons.
First, if you are trying to do something, whether or not people are making fun of it is no... |
18b9b070-c5c1-47ee-8b37-3516c755c4cf | trentmkelly/LessWrong-43k | LessWrong | On the Confusion between Inner and Outer Misalignment
Here’s my take on why the distinction between inner and outer-alignment frame is weird/unclear/ambiguous in some circumstances: My understanding is that these terms were originally used when talking about AGI. So outer alignment involved writing down a reward or ut... |
65dea091-899e-44fb-b703-64943db20302 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Rationality Reading Group (76-80)
Discussion article for the meetup : Rationality Reading Group (76-80)
WHEN: 24 August 2015 06:30:00PM (-0700)
WHERE: Paul G. Allen Center (185 Stevens Way, Seattle, WA) Room 503
Reading group for Yudkowsky's "Rationality: AI to Zombies", which is basically an organized and... |
9e25b62a-3380-4661-8cbb-6211fe9ffabe | trentmkelly/LessWrong-43k | LessWrong | Takeaways from calibration training
Summary. Thoughts arising from doing calibration training on ~500 questions, consisting of both trivia questions and mundane real life events.
Epistemic status: Confident on own experiences, uncertain on generalizability and usefulness to others.
Intro
I've practiced quantifying... |
f40a3fb3-3890-40ae-9f5a-f8131a468ca0 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | FDT is not directly comparable to CDT and EDT
In [*Functional Decision Theory: A New Theory of Instrumental Rationality*](https://arxiv.org/pdf/1710.05060.pdf) and [*Cheating Death in Damascus*](https://intelligence.org/files/DeathInDamascus.pdf), the authors go through a multitude of decision problems and compare th... |
d446f73e-bdc9-4752-9146-3ffe352bf7f3 | trentmkelly/LessWrong-43k | LessWrong | In Wikipedia — reading about Roko's basilisk causing "nervous breakdowns" ...
I recently came upon the story of "Roko's basilisk". Heretofore I'd not seen the topic. Reading about it in Wikipedia[1] I saw "Discussion of Roko's basilisk was banned on LessWrong for several years because Yudkowsky had stated that it ca... |
92459ab6-7c38-4d18-9970-eeae81a64e51 | trentmkelly/LessWrong-43k | LessWrong | I grade every NBA basketball game I watch based on enjoyability
It started as a spreadsheet, then I built a simple website to make it easier. It's like Letterboxd/Goodreads but for NBA games - find any game since the 1984-85 season, give it a letter grade, add notes, and create lists.
https://hooplog.io/
It's 100% f... |
858a8d53-d3a0-4f58-9841-c09559a6e9fd | trentmkelly/LessWrong-43k | LessWrong | Learned Helplessness
I stumbled upon this interesting little video. It links to many others on the same topic, which I found surprisingly interesting, and would like to discuss. What is it that makes one lose confidence when they see others succeed at the tasks they fail, to the point of being unable to preform tasks ... |
273d4eb1-b94b-49db-9243-9ab7c735260d | trentmkelly/LessWrong-43k | LessWrong | An overview of 11 proposals for building safe advanced AI
This is the blog post version of the paper by the same name. Special thanks to Kate Woolverton, Paul Christiano, Rohin Shah, Alex Turner, William Saunders, Beth Barnes, Abram Demski, Scott Garrabrant, Sam Eisenstat, and Tsvi Benson-Tilsen for providing helpful ... |
dc529a27-c407-46ab-b7cd-35828ef9e7d9 | trentmkelly/LessWrong-43k | LessWrong | Do You Really Want Effective Altruism?
There is something fundamentally different between the way nature solves complex problems and the way human beings currently do. When nature wants to solve problems beyond a certain level of complexity it uses networks. From whatever evidence we can find that remains of the billi... |
fe4d61fe-8cd7-42d3-a402-376b3c05a502 | trentmkelly/LessWrong-43k | LessWrong | Contingency: A Conceptual Tool from Evolutionary Biology for Alignment
-1. Motivation
As AI systems become increasingly powerful, the chance of them developing dangerous features becomes increasingly likely. A key concern of AI alignment researchers is that there are dangerous features, such as power-seeking or situat... |
f147c5f9-9d0a-47ff-9f9f-d1a9d500ff8f | trentmkelly/LessWrong-43k | LessWrong | AXRP Episode 24 - Superalignment with Jan Leike
YouTube link
Recently, OpenAI made a splash by announcing a new “Superalignment” team. Lead by Jan Leike and Ilya Sutskever, the team would consist of top researchers, attempting to solve alignment for superintelligent AIs in four years by figuring out how to build a tr... |
6a827175-a8f1-4b87-83d0-e97abb7f095f | trentmkelly/LessWrong-43k | LessWrong | Are the Sciences Better Than the Social Sciences For Training Rationalists?
A Wall Street Journal article by Harvard professor of government Harvey Mansfield claims that the social sciences and humanities are inferior to the sciences. The article implicitly urges undergraduates to major in science. From the article:... |
3ecd936a-ef11-4546-86af-a58cacfce4ff | trentmkelly/LessWrong-43k | LessWrong | Rationalist Wiki
Some (including myself) have suggested that a rationality wiki would be a useful supplement to this site. In the spirit of getting things done, I set one up here: http://rationality.tiddlyspot.com/ The password to save edits is omega.
The TiddlyWiki framework it uses is very lightweight and won't be ... |
065ca4f4-7012-4b6f-845a-00e7f373f5a6 | StampyAI/alignment-research-dataset/aisafety.info | AI Safety Info | What is "metaphilosophy" and how does it relate to AI safety?
[Metaphilosophy](https://www.alignmentforum.org/tag/meta-philosophy) is the study of how philosophy is developed. Wei Dai [introduced this idea](https://www.lesswrong.com/posts/MAhueZtNz5SnDPhsy/metaphilosophical-mysteries) as a way to think about AI safety... |
6c858fc9-4017-4ae5-a08f-7c4d48398b9f | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "So we say we know evolution is an alien god, which can do absolutely horrifying things to creatures. And surely we are aware that includes us, but how exactly does one internalize something like that? Something so at odds with default cultural intuitions. It may be just my mood tonight, but this short entry on the Wes... |
e95f26d7-5735-4616-8a1b-2ca31de875e1 | trentmkelly/LessWrong-43k | LessWrong | The O'Brien Technique
Schelling place for comments is here on LessWrong. |
755fae91-5fa3-4766-8637-399104249968 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] No License To Be Human
Today's post, No License To Be Human was originally published on 20 August 2008. A summary (taken from the LW wiki):
> We don't pull children off of train tracks because that's what humans do, or because we can consistently hold that it's better to pull children off of train trac... |
7ec38765-8882-49ab-93f8-97534d850a60 | trentmkelly/LessWrong-43k | LessWrong | Beware surprising and suspicious convergence
[Cross]
Imagine this:
> Oliver: … Thus we see that donating to the opera is the best way of promoting the arts.
>
> Eleanor: Okay, but I’m principally interested in improving human welfare.
>
> Oliver: Oh! Well I think it is also the case that donating to the opera is b... |
92611bba-59ab-4020-be86-4bd8d3e4697e | trentmkelly/LessWrong-43k | LessWrong | Will LLM agents become the first takeover-capable AGIs?
One of my takeaways from EA Global this year was that most alignment people aren't explicitly focused on LLM-based agents (LMAs)[1] as a route to takeover-capable AGI. I want to better understand this position, since I estimate this path to AGI as likely enough (... |
55bd953c-8ea5-4a77-8879-f1f4071d683e | trentmkelly/LessWrong-43k | LessWrong | Risk aversion and GPT-3
This post aims to build on the recent work by Binz and Schulz [1]. I did an experiment which looked at the preferences of GPT-3 at tacking a gamble versus some fixed amount (the expected value of the bet with a discount).
Repo: https://github.com/afiqhatta/gpt_risk_aversion
Question
* Does... |
d2ed49ed-0584-474a-a1a8-27705c9c145f | trentmkelly/LessWrong-43k | LessWrong | Tagging Open Call / Discussion Thread
You’ve listened to the LessWrong team talk about our new tagging feature for months. First a steady drip of we’re working on it, then announcements of various milestones like you can now filter Coronavirus in or out and anyone can create tags. Well, now, it's an open call for tag... |
660f17c2-6ee6-466d-865e-72657bbcd92a | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | What if memes are common in highly capable minds?
The meme-theoretic view of humans says: Memes are to humans as sailors are to ships in the age of sail.
If you want to predict where a ship will go, ask: Is it currently crewed by the French or the English? Is it crewed by merchants, pirates, or soldiers? These are t... |
dfae45ee-3352-4f49-9d2d-cd1e07c0fe5f | trentmkelly/LessWrong-43k | LessWrong | Weekly LW Meetups
This summary was posted to LW Main on July 31st. The following week's summary is here.
New meetups (or meetups with a hiatus of more than a year) are happening in:
* [Indianapolis] Gen Con: Applied Game Theory: 01 August 2015 02:00PM
* Kyiv, Ukraine: 02 August 2015 06:30PM
Irregularly scheduled ... |
e02ee002-60be-482e-9158-768365b1d1cb | trentmkelly/LessWrong-43k | LessWrong | Meetup : Berlin Meetup
Discussion article for the meetup : Berlin Meetup
WHEN: 04 September 2012 07:30:00PM (+0200)
WHERE: Spreegold, Hufelandstraße 20, 10407 Berlin
We're meeting for dinner at http://www.spreegold.com . There'll be a sign. Please leave a comment if you're planning to attend.
Discussion article f... |
09057480-924f-4490-abdb-07d79bdd60fe | trentmkelly/LessWrong-43k | LessWrong | What is the subjective experience of free will for agents?
Jessica recently wrote about difficulties with physicalist accounts of the world and alternatives to logical counterfactuals. In my recent post about the deconfusing human values research agenda, Charlie left a comment highlighting that my current model depend... |
c968b814-0684-4496-adf5-6342f5244f56 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Salt Lake, UT: Lightning Talks
Discussion article for the meetup : Salt Lake, UT: Lightning Talks
WHEN: 09 August 2014 04:45:00PM (-0600)
WHERE: Steamhead Cafe, 2470 S Main St, Salt Lake City, Utah 84115
The irony of hosting a rationalist meetup by a witch's store amuses me. This actually takes place right... |
7452f8a7-7e5a-4d25-8821-16ecb30e3f1c | trentmkelly/LessWrong-43k | LessWrong | The economy of weirdness
It is often said that you should spend your weirdness budget wisely. You should wear a gender-appropriate suit, and follow culture-appropriate sports, and use good grammar, and be non-specifically spiritual, and support moderate policies, and not have any tattoos around either of your eyes. An... |
60b439f0-cce5-47fb-a763-c3ccd6332d23 | trentmkelly/LessWrong-43k | LessWrong | How do you organise your reading?
Maybe this is just a personal problem, but I frequently find myself with just an absolute mess of things to read. Blog posts over several different sites, PDFs, paper books, podcasts and other content I intend to consume. I suspect a significant proportion of it falls through the crac... |
814f56ae-fa99-4c68-b99b-b8c182b89115 | trentmkelly/LessWrong-43k | LessWrong | SGD Understood through Probability Current
My previous post about SGD was an intro to this model. That post concerned a model of a loss landscape on two "datapoints". In this post I attempt to build a new model of SGD and validate it, with mixed success, but it is sort of interesting.
Gradient Variance
We could mode... |
25fc12c4-b149-4735-a06f-7779be99effb | StampyAI/alignment-research-dataset/aisafety.info | AI Safety Info | What is a shoggoth?
[Shoggoths](https://en.wikipedia.org/wiki/Shoggoth) are alien creatures from [H. P. Lovecraft’s Cthulhu Mythos](https://en.wikipedia.org/wiki/Cthulhu_Mythos) that are so incomprehensible they would drive humans insane if we tried to understand them. The ‘shoggoth with a human mask’ has [emerged as ... |
2ae3d19e-08f2-468d-af2e-f9f4496e7002 | trentmkelly/LessWrong-43k | LessWrong | On not getting a job as an option
This was originally a comment to VipulNaik's recent indagations about the academic lifestyle versus the job lifestyle. Instead of calling it lifestyle he called them career options, but I'm taking a different emphasis here on purpose.
Due to information hazards risks, I recommend tha... |
41568036-7783-4faa-b29d-8d1725c56e03 | trentmkelly/LessWrong-43k | LessWrong | Roleplaying As Yourself
(This is a basic intuition pump I've found helpful in making decisions, and maybe you'll like it too.)
For all its shortcomings, I think there was something quite useful about the "What Would Jesus Do?" meme within the Christian framework. Of course it's not a very sophisticated ethical guide,... |
c1588906-efe4-4b00-9316-a15702252fa1 | trentmkelly/LessWrong-43k | LessWrong | 2014 Less Wrong Census/Survey
It's that time of year again.
If you are reading this post and self-identify as a LWer, then you are the target population for the Less Wrong Census/Survey. Please take it. Doesn't matter if you don't post much. Doesn't matter if you're a lurker. Take the survey.
This year's census cont... |
1fdbf4ed-5e5b-4ef0-94e9-f0a03b273b8a | trentmkelly/LessWrong-43k | LessWrong | Stability of optimal predictor schemes under a broader class of reductions
We define a new class of reductions which preserve optimal predictor schemes, generalizing the previous notion of pseudo-invertible reductions. These are reductions that preserve the estimation problem on average but allow for large variance.
... |
319e36cf-9f4d-4732-bfe3-0956359b1fe4 | trentmkelly/LessWrong-43k | LessWrong | [Valence series] 5. “Valence Disorders” in Mental Health & Personality
5.1 Post summary / Table of contents
Part of the Valence series.
Here in the final post of the Valence series, I will discuss how valence might shed light on three phenomena in mental health and personality: depression, mania, and narcissistic per... |
a2190732-7dd7-467e-856f-8f2e24c4c743 | trentmkelly/LessWrong-43k | LessWrong | Some Unorthodox Ways To Achieve High GDP Growth
GDP growth, as traditionally calculated, is a weird metric. People interpret it as measuring “economic growth”, but… well, think about electronics. Electronics which would have cost millions of dollars (or more) in 1984 are now commonplace, everyone carries them around i... |
04e078f9-4ffe-4d83-9df1-9ee09af22eb9 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Learning the prior and generalization
*This post is a response to Paul Christiano's post “[Learning the prior](https://www.alignmentforum.org/posts/SL9mKhgdmDKXmxwE4/learning-the-prior).”*
The generalization problem
==========================
Generally, when we train models, we often end up deploying them in situa... |
7855a947-8dd2-4781-a851-6c70eb088f16 | trentmkelly/LessWrong-43k | LessWrong | Where to Intervene in a Human?
The "What is Rationality?" page on the new CFAR website contains an illuminating story about Intel:
> Semiconductor giant Intel was originally a memory chip manufacturer. But by 1985, memory chips had been losing them money for years. Co-founders Andy Grove (CEO) and Gordon Moore met to... |
f20c3afe-9fdd-48aa-b1cd-338ae468599e | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Encultured AI Pre-planning, Part 2: Providing a Service
*Also available on the* [*EA Forum*](https://forum.effectivealtruism.org/posts/MWWZQ8C655iT9zzRd/encultured-ai-part-2-providing-a-service)*.*
*Preceded by:* [*Encultured AI Pre-planning, Part 1: Enabling New Benchmarks*](https://www.lesswrong.com/posts/AR6mfyd... |
42fd5a6f-acf9-449e-ba81-86e07f9a5afb | trentmkelly/LessWrong-43k | LessWrong | MIRI Fundraiser: Why now matters
Our summer fundraiser is ongoing. In the meantime, we're writing a number of blog posts to explain what we're doing and why, and to answer a number of common questions. Previous posts in the series are listed at the above link.
----------------------------------------
I'm often asked... |
2aa15cd8-f880-4e64-986a-8763363d24da | trentmkelly/LessWrong-43k | LessWrong | Why AI may not foom
Summary
* There's a decent chance that the intelligence of a self-improving AGI will grow in a relatively smooth exponential or sub-exponential way, not super-exponentially or with large jump discontinuities.
* If this is the case, then an AGI whose effective intelligence matched that of the worl... |
f347fe91-c6f2-4a48-b0d8-b567c8418b87 | trentmkelly/LessWrong-43k | LessWrong | New Year's Prediction Thread (2014)
It's time to look back to see what was predicted a year ago and how successfully it was.
But even more, it's time for the fresh predictions for the following year, 2014. |
2776ee32-8d1d-43ac-bd1d-62506102555b | trentmkelly/LessWrong-43k | LessWrong | Sunday Meetup: Workshop on Online Surveys with Spencer Greenberg
Alternate title: "Geez LessWrong peeps do more empiricism please".
At this Sunday's LW meetup, special guest Spencer Greenberg will be running a workshop on how to run a psychology study using online tools.
The LessWrong community raises a lot of int... |
3586597a-f513-4d96-836d-f765aa3f3c14 | trentmkelly/LessWrong-43k | LessWrong | Karma fluctuations?
I see karma on posts fluctuating (in particular going down) more than I would expect coming from other vote-based websites. Is downvoting really used here for posts that are not spam or trolling? Or do people just change their minds a lot?
The FAQ has: We encourage people to vote such that upvote ... |
dda299b7-b9bc-461f-9b96-60ad54c1c57c | trentmkelly/LessWrong-43k | LessWrong | [Alignment] Is there a census on who's working on what?
I'm new to the Rationality / EA community, and I've been getting the sense that the best use of my skills and time is to try to contribute to alignment.
Currently, my focus has been guided by top posts / most often mentions of names, here on LW. E.g. "I see John... |
51b84dfe-a832-4f7c-8cde-14966994b9ca | trentmkelly/LessWrong-43k | LessWrong | Meetup : West LA—R:AZ Part E, Overly Convenient Excuses
Discussion article for the meetup : West LA—R:AZ Part E, Overly Convenient Excuses
WHEN: 06 May 2015 07:00:00PM (-0700)
WHERE: 11066 Santa Monica Blvd, Los Angeles, CA
How to Find Us: Go into this Del Taco. We will be in the back room if possible.
Parking is ... |
e114b803-3bf6-43d6-b27c-f7e80fb12310 | trentmkelly/LessWrong-43k | LessWrong | Change utility, reduce extortion
Crossposted at the Intelligent Agents Forum.
EDIT: This method is not intended to solve extortion, just to remove the likelihood of extremely terrible outcomes (and slightly reduce the vulnerability to extortion).
A full solution to the extortion problem is sorely elusive. However, t... |
6f579c37-ffb6-4eff-8d1c-e52570c15c9e | trentmkelly/LessWrong-43k | LessWrong | PhD Position: AI Interpretability in Berlin, Germany
We are seeking a highly motivated and skilled research assistant to join our Research Team at IOL Lab at Zuse Institute Berlin. The successful candidate will work closely with our team on cutting-edge research projects in the field of Numerical Optimisation and Expl... |
9fc0641b-9792-4d37-acf9-33c687f4a261 | trentmkelly/LessWrong-43k | LessWrong | Worries about latent reasoning in LLMs
When working through a problem, OpenAI's o1 model will write a chain-of-thought (CoT) in English. This CoT reasoning is human-interpretable by default, and I think that this is hugely valuable. Assuming we can ensure that these thoughts are faithful to the model's true reasoning,... |
782556c8-5833-4e76-8fb4-105c2f65e318 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] But There's Still A Chance, Right?
Today's post, But There's Still A Chance, Right? was originally published on 06 January 2008. A summary (taken from the LW wiki):
> Sometimes, you calculate the probability of a certain event and find that the number is so unbelievably small that your brain really can... |
6ea87992-d87d-4d74-8c36-2c634b579aea | trentmkelly/LessWrong-43k | LessWrong | Weekly LW Meetups
This summary was posted to LW Main on June 17th. The following week's summary is here.
Irregularly scheduled Less Wrong meetups are taking place in:
* Baltimore weekly meetup: 19 June 2016 08:00PM
* European Community Weekend: 02 September 2016 03:35PM
* San Antonio Meetup: 19 June 2016 02:00PM
... |
6aa0c2b5-dcdb-4ed6-91a0-0944c2bb19f0 | trentmkelly/LessWrong-43k | LessWrong | Side-channels: input versus output
This is a brief post arguing that, although "side-channels are inevitable" is pretty good common advice, actually, you can prevent attackers inside a computation from learning about what's outside.
We can prevent a task-specific AI from learning any particular facts about, say, huma... |
650b753c-9630-4fce-926e-b323c650965d | trentmkelly/LessWrong-43k | LessWrong | A philosophy professor elicits college students' reactions to Less Wrong
Link.
Given the positive reactions, I think the professor seeded them with a positive impression of the site's content. |
d75cfabf-cf2a-4f14-9908-19c103cc8a06 | trentmkelly/LessWrong-43k | LessWrong | Self-study ideas for micro-projects in "abstract" subjects?
Music and the arts, programming, and cooking have a wonderful property in common. They are all extremely compatible with micro-projects. By this, I mean that it's pretty easy for an artist, coder, or cook of any skill level to find a real-world, interesting p... |
ee6a07e1-0cad-431b-a2cf-a5a1eb928093 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Deontology for Consequentialists
*Consequentialists see morality through consequence-colored lenses. I attempt to prise apart the two concepts to help consequentialists understand what deontologists are talking about.*
Consequentialism1 is built around a group of variations on the following basic assumption:
* Th... |
7441bd23-c341-4c04-b708-ba7602b410f5 | trentmkelly/LessWrong-43k | LessWrong | AI #100: Meet the New Boss
Break time is over, it would seem, now that the new administration is in town.
This week we got r1, DeepSeek’s new reasoning model, which is now my go-to first choice for a large percentage of queries. The claim that this was the most important thing to happen on January 20, 2025 was at lea... |
ad348c65-9244-43d5-88ae-5d3ed2f01bc7 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "Related to: Dissolving the Question, Words as Hidden Inferences. In what sense is the world “real”? What are we asking, when we ask that question?
I don’t know. But G. Polya recommends that when facing a difficult problem, one look for similar but easier problems that one can solve as warm-ups. I would like to do one ... |
8b6e4a0b-68e4-418c-9cbd-b440eab52bf3 | trentmkelly/LessWrong-43k | LessWrong | Informed Oversight through Generalizing Explanations
|
384a120c-0640-4b89-9506-110002c8af88 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | In a multipolar scenario, how do people expect systems to be trained to interact with systems developed by other labs?
I haven't seen much discussion of this, but it seems like an important factor in how well AI systems deployed by actors with different goals manage to avoid conflict (cf. my discussion of equilibriu... |
485ae1ee-3944-4080-a309-558c0fb5f665 | trentmkelly/LessWrong-43k | LessWrong | The rational way to name rivers
I just read this in the Wikipedia article on the Mattaponi River and it really tickled me. If only all language were so rational!
> * The Mat River and the Ta River join in Spotsylvania County to form the Matta River;
> * The Po River and the Ni River join in Caroline County to form ... |
75e92ad3-afe8-435b-ab22-b79854378fe6 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | More Christiano, Cotra, and Yudkowsky on AI progress
This post is a transcript of a discussion between Paul Christiano, Ajeya Cotra, and Eliezer Yudkowsky (with some comments from Rob Bensinger, Richard Ngo, and Carl Shulman), continuing from [1](https://www.lesswrong.com/posts/vwLxd6hhFvPbvKmBH/yudkowsky-and-christia... |
02570d5a-5a41-41e7-9ec5-aec104da636a | trentmkelly/LessWrong-43k | LessWrong | Charbel-Raphaël and Lucius discuss interpretability
The following is a transcript of a public discussion between Charbel-Raphaël Segerie and Lucius Bushnaq that took place on 23 September during LessWrong Community Weekend 2023. I edited the transcript for clarity.
Transcript
Mateusz: Last month, Charbel published t... |
62b55f46-268e-49e6-a895-637a2e5df21d | trentmkelly/LessWrong-43k | LessWrong | Deconstructing 321 Voting
This is a quick note about an alternate voting system which occurred to me while talking to Scott Garrabrant about 321 voting.
Above everything else, you should know that 321 voting is very good. It rates very highly in VSE, and is quite difficult to warp with strategic voting.
The basic id... |
5dd26e7f-6083-428e-b827-3563adc130e9 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Grasping Slippery Things
Today's post, Grasping Slippery Things was originally published on 17 June 2008. A summary (taken from the LW wiki):
> An illustration of a few ways that trying to perform reductionism can go wrong.
Discuss the post here (rather than in the comments to the original post).
Th... |
9f2d31fa-addc-4787-a7cd-f568c6cbc971 | StampyAI/alignment-research-dataset/arbital | Arbital | User manipulation
If there's anything an AGI wants whose achievement involves steps that interact with the AGI's programmers or users, then by default, the AGI will have an [instrumental incentive](https://arbital.com/p/10k) to optimize the programmers/users in the course of achieving its goal. If the AGI wants to se... |
77340ca7-2aa2-4fef-b5b0-23e430164d03 | trentmkelly/LessWrong-43k | LessWrong | The Gemini 1.5 Report
This post goes over the extensive report Google put out on Gemini 1.5.
There are no important surprises. Both Gemini Pro 1.5 and Gemini Flash are ‘highly capable multimodal models incorporating a novel mixture-of-experts architecture’ and various other improvements. They are solid models with so... |
2a86fb8c-c4d5-47b4-a4e3-67230717e2a5 | trentmkelly/LessWrong-43k | LessWrong | Exposition and guidance by analogy
(Below I attempt to reproduce a chart I can't [?] embed, from Models and Analogies in Science by Mary Hesse; I don't know how it will display on all screens, but the source is this image at the top of the cross-post version here.)
At an elementary level we can set up the following c... |
30bef90d-8585-4d42-a48c-249f01eb836a | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Proposal: Scaling laws for RL generalization
*In this post, we (Alexander Meulemans, David Lindner, Florian Dorner) propose to study scaling laws relating the generalization capability of Reinforcement Learning (RL) agents to problem complexity, inspired by DeepMind’s recent success in training agents that perform wel... |
81432d12-c1b4-426a-a2f3-5ee10f122ed6 | trentmkelly/LessWrong-43k | LessWrong | Experiences are friends
Products tend to be less satisfying than experiences. This is old news. Psyblog elaborates on six reasons from a set of (hard to access) recent experiments, which add up to this:
We compare products more than experiences, and since products are doomed to not be the best we could ever have got,... |
46fea306-7a39-4fa9-88eb-cd58eb31e33c | StampyAI/alignment-research-dataset/blogs | Blogs | 2014 Winter Matching Challenge!
Update: **we have finished the matching challenge!** Thanks everyone! The original post is below.
Thanks to the generosity of Peter Thiel,[1](https://intelligence.org/2014/12/01/2014-winter-matching-ch... |
4e99056b-c7de-446c-9c9d-b4dd108a5447 | trentmkelly/LessWrong-43k | LessWrong | Prizes for the 2020 Review
The third annual review is complete. We spent two months nominating, reviewing, and voting on the best posts of 2020. After lots of thoughtful evaluation, it's time to award some prizes!
As described in the voting results post, there are two prize pools for posts. Lightcone Infrastructure ... |
d7c534a9-b753-413d-b7ce-3122f23624f1 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | For the past, in some ways only, we are moral degenerates
Have human values improved over the last few centuries? Or is it just that current human values are naturally closer to our (current) human values and so we think that there's been moral progress towards us?
If we project out in the future, the first scenario... |
44dbe926-94e4-479f-aeaf-9707526679b5 | trentmkelly/LessWrong-43k | LessWrong | FHI postdoc at Oxford
The Future of Humanity Institute at the University of Oxford is looking to fill a postdoctoral research fellowship in interdisciplinary science or philosophy. Current work areas include global catastrophic risks, probabilistic methodology and applied epistemology and rationality, impacts of futu... |
6f4c41c4-c235-4129-9637-b240be5a2f59 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | [Link Post] Interesting shallow round-up of reasons to be skeptical that transformative AI or explosive economic growth are coming soon
Here's a blogpost by [Arjun Ramani and Zhengdong Wang](https://thegradient.pub/author/arjun-ramani-and-zhengdong-wang/)\* rounding up some reasons for skepticism about transformative... |
704b5ecf-6059-4218-bd68-8f8a5bb780e7 | StampyAI/alignment-research-dataset/special_docs | Other | The truth behind the myth of the folk theorem.
The truth behind the myth of the Folk theorem[☆](#aep-article-footnote-id1)
===========================================================================
Abstract
--------
We study the problem of computing an *ϵ*-Nash equilibrium in repeated games. Earlier work by [Borgs ... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.